Estimating pore pressure in tight sandstone gas reservoirs: A comprehensive approach integrating rock physics models and deep neural networks

Abstract

Pore pressure serves as an important driving power for subsurface fluid migration and therefore has a significant impact on gas accumulation and enrichment in tight sandstone reservoirs. Tight gas is typically produced in overpressure regions, where pressure coefficients are notably elevated. Thus, it is crucial to establish an effective methodology for precise pore pressure estimation. This study introduces an approach to improve pore pressure prediction by incorporating rock physical modeling and deep neural networks (DNNs) into the classical Eaton method. Compared to conventional techniques relying on empirical correlations between pressure coefficients and elastic properties, the proposed method considers the influence of porosity, fluids, and lithology, which could enhance reliability in pore pressure prediction. Meanwhile, a prediction model is developed using logging data and DNNs to estimate mineralogical volumetric fractions based on elastic properties. This prediction model allows improved estimation of rock matrix elastic properties using seismic-inverted data, which is crucial for estimating normal compaction velocity to extend pore pressure prediction from individual boreholes to the whole study area. Real data applications demonstrate that the predicted pressure coefficients derived from seismic data using the method presented in this paper align well with the gas enrichment estimated in previous studies for the tight sandstone reservoirs. Furthermore, regions with high values of pressure coefficients correspond to high gas content. These findings validate the effectiveness of the proposed methodology, which can provide valuable insights for identifying potential tight sandstone reservoirs.